ICM 2026 Public Lecture - Geordie Williamson

Value of the Information & Strength of the Argument

The lecture provides valuable insights into the intersection of AI and mathematics, combining historical context with recent developments. Williamson’s argumentation is solid, drawing on his own research and well-known examples. He effectively explains complex concepts like next-token prediction and scaling laws in an accessible manner. He also presents a balanced view, acknowledging both the potential and the challenges of AI in mathematics. The anecdotal evidence from his own work, while compelling, is presented with appropriate caveats about reporting bias.

Scientific Rigor, Source Quality, Title Accuracy

Williamson demonstrates scientific rigor by referencing key papers and historical figures, such as Turing’s ‘Intelligent Machinery’ and Shannon’s ‘A Mathematical Theory of Communication’. He also mentions specific works like Kaplan et al.’s scaling laws paper. The title accurately reflects the content. The lecture is well-structured and the sources are credible. However, as a public lecture, it lacks detailed citations and some claims are based on personal experience rather than published research.

The title accurately reflects the content: a public lecture by Geordie Williamson at ICM 2026, focusing on AI and mathematics.

8/10

Lecture by a leading mathematician with deep expertise in representation theory and AI applications in mathematics. The content is well-structured, historically grounded, and includes references to key papers and recent developments. However, it is a public lecture with limited technical depth and some anecdotal elements.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction of Geordie Williamson by the host.
• 02:00Williamson introduces the concept of mathematics as a long conversation.
• 05:00Discussion of Alan Turing's paper 'Intelligent Machinery' and his list of areas for AI.
• 10:00Reference to Timothy Gowers' 1999 article imagining a conversation between a mathematician and a computer.
• 15:00Explanation of Claude Shannon's work on approximating English and next-token prediction.
• 20:00Introduction to transformers and their training via gradient descent.
• 25:00Discussion of scaling laws and their implications.
• 30:00Visualization of the emergent geometry inside a transformer.
• 35:00Example of using an LLM to find large hypercubes in a graph.
• 40:00Mention of the counterexample to the Erdős unit distance conjecture.
• 45:00Concerns about AI's impact on the next generation of mathematicians.

• Intelligent Machinery — Turing's paper on AI, mentioned as the first paper on artificial intelligence.
• A Mathematical Theory of Communication — Shannon's paper, discussed in the context of next-token prediction.
• Scaling Laws for Neural Language Models — Kaplan et al.'s paper on scaling laws, referenced in the lecture.

• Scaling Laws for Neural Language Models — The paper by Kaplan et al. that Williamson references, supporting the scaling laws discussion.